I think it comes from separating domain experts from writing software themselves. There was a lot of great software for different niche things that were written

I think it comes from separating domain experts from writing software themselves. There was a lot of great software for different niche things that were written by people who were some what experts in these niche things that happened to write code as well. They knew what calculations people in this field would be doing, and simply wrote the function and wrapped it in some tool. Immediately useful software. A lot of "academic" code that is pilloried by "real" software engineers is actually a great example of this. Is it the most performant? No. Could a random person off the street make use of this? No. But for those in the field, they know exactly what this tool is conceptually doing even if they don't know how it is made exactly. It is like a very specialized tool for a very specialized tradesman. What is interesting about the shift away from utilitarian programming is that these "thought leaders" and other middlemen now find themselves in positions to set the narrative essentially thanks to monopolizing the space, change the very meaning of work to suit the software they happen to peddle rather than the other way around. We saw this with enterprise software and now we see this with AI tooling shoehorned into that enterprise software.